Doc-Id Reader (Brazilian documents)
Daniel Theisges dos Santos
Unknown
- 0 Collaborators
Recognizing identity documents is an important activity that can be used in a variety of applications. This project aimed to track, identify and recognize the texts present in Brazilian identity documents. ...learn more
Project status: Concept
Groups
Student Developers for AI,
DeepLearning,
Movidius™ Neural Compute Group
Intel Technologies
Intel Opt ML/DL Framework,
Movidius NCS,
OpenVINO
Overview / Usage
The initial objective of this work is to recognize and track two types of Brazilian documents: Id document and driver's license. For this, a transfer learning of the YOLO neural network implementation will be used, available in the Darkflow project. See the video demo.
Methodology / Approach
The methodology is divided into two stages:
- Scraping and identification of documents: Transfer learning through the implementation of the YOLO neural network.
- Text and OCR Scheduling: Use implementations already available in OpenCV.
Technologies Used
Implementation in tensorflow for identification and tracking of objects.
The inference and classification steps were performed in CPU and VPU. This project uses OpenVINO and Movidius to detect objects in real time.